RE: Blog #5: Characterising Thermistors – Measuring Resistance Is Not So Easy!

Andrew J — Fri, 12 Aug 2022 10:01:05 GMT

As ever, really useful insight Gough.

RE: Blog #5: Characterising Thermistors – Measuring Resistance Is Not So Easy!

DAB — Thu, 11 Aug 2022 20:24:51 GMT

Very good analysis Gough.

RE: Blog #5: Characterising Thermistors – Measuring Resistance Is Not So Easy!

dougw — Thu, 11 Aug 2022 14:09:43 GMT

Nice work Gough. It is interesting to see how meters measure resistance.

RE: Blog #5: Characterising Thermistors – Measuring Resistance Is Not So Easy!

genebren — Thu, 11 Aug 2022 13:57:55 GMT

Great blog with plenty of thought provoking analysis. I really appreciate the depth and breath approach that you take when digging a subject. Well done!

RE: Blog #5: Characterising Thermistors – Measuring Resistance Is Not So Easy!

michaelkellett — Thu, 11 Aug 2022 10:36:37 GMT

More good stuff Gough.

Just a quick note about oversampling.

(I have enough scars to make it a long note but it gets a little esoteric.)

If you have a perfectly stable and noise free system then simple oversampling offers no benefit at all.

Every sample will be exactly the same and just one will give the same result as averaging a million.

If you add some dither to the system (an offset added to the signal which varies from sample to sample) then averaging multiple samples starts to work.

There is often enough noise in a real system to act as dither, but incedental noise may well not be random. Ideal dither will be a signal ramping up and down over a period equal to number of samples averaged (or a sub multiple of it).

Of course real ADCs are non linear, a very few will offer linearity better than 1LSB but most (and especially the ones in micro-controllers a much worse.)

You might expect that the biggest non-linear jump in an ADC would occur at the transition of the msb, and indeed it often does, but in real 12 bit ADC significant jumps can occur at the transitions of bit 11,10,9,8 and even lower.

It is very difficult to eliminate these non linear effects by averaging and random noise dithering. (Probably very difficult using non random dithering too.)

I could go on, but suffice to say that oversampling will help get rid of noise but won't (without a lot of trickery) make a 10 bit ADC with 2 lsb non linearity work as well as a a 12 bit ADC with 2 lsb non linearity.